In this notebook, I use a causal machine learning estimator, i.e., multi-armed causal random forest with augmented inverse propensity score weights (Athey et al 2019), to estimate conditional average treatment effects (CATES) for agronomic practices. These CATEs are estimated for each individual farm thereby providing personalized estimates of the potential effectiveness of the practices. I then use a debiased robust estimator in a policy tree optimization (Athey and Wager 2021) to generate optimal recommendations in the form of agronomic practices that maximize potential yield gains.
load("CSISA_KVK_Public_Workspace.RData")
CSISA_KVKestim=CSISA_KVK
table(CSISA_KVKestim$SowingSchedule)
T1 T2 T3 T4 T5
408 858 806 881 695 CSISA_KVKestim$SowingSchedule=ordered(CSISA_KVKestim$SowingSchedule,levels=c("T5","T4","T3","T2","T1"))
#Create dummies for some categorical variables
library(fastDummies)
CSISA_KVKestim <- fastDummies::dummy_cols(CSISA_KVKestim, select_columns =c("VarietyClass","SoilType","CropEstablishment","Year","District"))# Bar graphs showing percentage of farmers adopting these practices
library(tidyverse)
library(ggplot2)
bar_chart=function(dat,var){ dat|> drop_na({{var}})|> mutate({{var}}:=factor({{var}})|>fct_infreq())|> ggplot()+ geom_bar(aes(y={{var}}),fill="dodgerblue4")+ theme_minimal(base_size = 16) }
sow_plot=bar_chart(CSISA_KVKestim,SowingSchedule)+labs(y="Sowing dates")
sow_plot
library(ggpubr)
library(tidyverse)
#Sowing dates
SowingDate_Options_Errorplot= CSISA_KVKestim%>%
drop_na(SowingSchedule) %>%
ggerrorplot(x = "SowingSchedule", y = "GrainYield",add = "mean", error.plot = "errorbar", color="steelblue", ggtheme=theme_bw())+
labs(x="Sowing date options",y="Wheat yield (t/ha)")+
theme_minimal(base_size = 16)+
coord_flip() Warning: The `fun.y` argument of `stat_summary()` is deprecated as of ggplot2 3.3.0.
i Please use the `fun` argument instead.
i The deprecated feature was likely used in the ggpubr package.
Please report the issue at <https://github.com/kassambara/ggpubr/issues>.Warning: The `fun.ymin` argument of `stat_summary()` is deprecated as of ggplot2 3.3.0.
i Please use the `fun.min` argument instead.
i The deprecated feature was likely used in the ggpubr package.
Please report the issue at <https://github.com/kassambara/ggpubr/issues>.Warning: The `fun.ymax` argument of `stat_summary()` is deprecated as of ggplot2 3.3.0.
i Please use the `fun.max` argument instead.
i The deprecated feature was likely used in the ggpubr package.
Please report the issue at <https://github.com/kassambara/ggpubr/issues>.SowingDate_Options_Errorplot 
library(fBasics)
summ_stats <- fBasics::basicStats(CSISA_KVKestim[,c("GrainYield","VarietyClass_LDV","SoilType_Heavy","SoilType_Medium","SoilType_Low","CropEstablishment_CT","CropEstablishment_CT-line","CropEstablishment_ZT")])
summ_stats <- as.data.frame(t(summ_stats))
# Rename some of the columns for convenience
summ_stats <- summ_stats[c("Mean", "Stdev", "Minimum", "1. Quartile", "Median", "3. Quartile", "Maximum")] %>%
rename("Lower quartile" = '1. Quartile', "Upper quartile"= "3. Quartile")
summ_stats Mean Stdev Minimum Lower quartile Median
GrainYield 4.067177 1.035497 1.39 3.23 4.06
VarietyClass_LDV 0.710526 0.453580 0.00 0.00 1.00
SoilType_Heavy 0.214090 0.410246 0.00 0.00 0.00
SoilType_Medium 0.763980 0.424693 0.00 1.00 1.00
SoilType_Low 0.021930 0.146475 0.00 0.00 0.00
CropEstablishment_CT 0.020285 0.140993 0.00 0.00 0.00
CropEstablishment_CT.line 0.004386 0.066090 0.00 0.00 0.00
CropEstablishment_ZT 0.975329 0.155142 0.00 1.00 1.00
Upper quartile Maximum
GrainYield 4.84 6.74
VarietyClass_LDV 1.00 1.00
SoilType_Heavy 0.00 1.00
SoilType_Medium 1.00 1.00
SoilType_Low 0.00 1.00
CropEstablishment_CT 0.00 1.00
CropEstablishment_CT.line 0.00 1.00
CropEstablishment_ZT 1.00 1.00library(grf)
library(policytree)
CSISA_KVKestim_sow=subset(CSISA_KVKestim, select=c("SowingSchedule","GrainYield","VarietyClass_LDV","SoilType_Heavy","SoilType_Medium","SoilType_Low","CropEstablishment_CT","CropEstablishment_CT-line","CropEstablishment_ZT","wc2.1_30s_elev","nitrogen_0.5cm","sand_0.5cm", "soc_5.15cm","Latitude","Longitude"))
library(tidyr)
CSISA_KVKestim_sow=CSISA_KVKestim_sow %>% drop_na()
Y_cf_sowing=as.vector(CSISA_KVKestim_sow$GrainYield)
## Causal random forest -----------------
X_cf_sowing=subset(CSISA_KVKestim_sow, select=c("VarietyClass_LDV","SoilType_Heavy","SoilType_Medium","SoilType_Low","CropEstablishment_CT","CropEstablishment_CT-line","CropEstablishment_ZT","wc2.1_30s_elev","nitrogen_0.5cm","sand_0.5cm", "soc_5.15cm","Latitude","Longitude"))
W_cf_sowing <- as.factor(CSISA_KVKestim_sow$SowingSchedule)
W.multi_sowing.forest <- probability_forest(X_cf_sowing, W_cf_sowing,
equalize.cluster.weights = FALSE,
seed = 2
)
W.hat.multi.all_sowing <- predict(W.multi_sowing.forest, estimate.variance = TRUE)$predictions
Y.multi_sowing.forest <- regression_forest(X_cf_sowing, Y_cf_sowing,
equalize.cluster.weights = FALSE,
seed = 2
)
print(Y.multi_sowing.forest)GRF forest object of type regression_forest
Number of trees: 2000
Number of training samples: 3628
Variable importance:
1 2 3 4 5 6 7 8 9 10 11 12 13
0.641 0.002 0.001 0.000 0.002 0.000 0.001 0.101 0.008 0.009 0.040 0.036 0.158 varimp.multi_sowing <- variable_importance(Y.multi_sowing.forest)
Y.hat.multi.all_sowing <- predict(Y.multi_sowing.forest, estimate.variance = TRUE)$predictions
multi_sowing.forest <- multi_arm_causal_forest(X = X_cf_sowing, Y = Y_cf_sowing, W = W_cf_sowing ,W.hat=W.hat.multi.all_sowing,Y.hat=Y.hat.multi.all_sowing,seed=2)
varimp.multi_sowing_cf <- variable_importance(multi_sowing.forest)
multi_sowing_ate=average_treatment_effect(multi_sowing.forest, method="AIPW")Warning in get_scores.multi_arm_causal_forest(forest, subset = subset, debiasing.weights = debiasing.weights, : Estimated treatment propensities take values very close to 0 or 1 meaning some estimates may not be well identified. In particular, the minimum propensity estimates for each arm is
T5: 0 T4: 0.004 T3: 0.002 T2: 0 T1: 0
and the maximum is
T5: 0.893 T4: 0.873 T3: 0.834 T2: 0.922 T1: 0.941.multi_sowing_ate estimate std.err contrast outcome
T4 - T5 0.3666129 0.04368794 T4 - T5 Y.1
T3 - T5 0.8570305 0.04010667 T3 - T5 Y.1
T2 - T5 0.8089736 0.22419766 T2 - T5 Y.1
T1 - T5 1.6852136 0.25117241 T1 - T5 Y.1varimp.multi_sowing_cf <- variable_importance(multi_sowing.forest)
vars_sowing=c("VarietyClass_LDV","SoilType_Heavy","SoilType_Medium","SoilType_Low","CropEstablishment_CT","CropEstablishment_CT-line","CropEstablishment_ZT","wc2.1_30s_elev","nitrogen_0.5cm","sand_0.5cm", "soc_5.15cm","Latitude","Longitude")
## variable importance plot ----------------------------------------------------
varimpvars_sowing=as.data.frame(cbind(varimp.multi_sowing_cf,vars_sowing))
names(varimpvars_sowing)[1]="Variableimportance_sowing"
varimpvars_sowing$Variableimportance_sowing=formatC(varimpvars_sowing$Variableimportance_sowing, digits = 2, format = "f")
varimpvars_sowing$Variableimportance_sowing=as.numeric(varimpvars_sowing$Variableimportance_sowing)
varimpplotRF_sowing=ggplot(varimpvars_sowing,aes(x=reorder(vars_sowing,Variableimportance_sowing),y=Variableimportance_sowing))+
geom_jitter(color="steelblue")+
coord_flip()+
labs(x="Variables",y="Variable importance")
previous_theme <- theme_set(theme_bw(base_size = 16))
varimpplotRF_sowing
# Policy tree --------------------------------------
DR.scores_sowing <- double_robust_scores(multi_sowing.forest)
tr_sowing <- policy_tree(X_cf_sowing, DR.scores_sowing, depth = 2)
plot(tr_sowing)tr_sowing3 <- hybrid_policy_tree(X_cf_sowing, DR.scores_sowing, depth = 3)
tr_sowing3policy_tree object
Tree depth: 3
Actions: 1: T5 2: T4 3: T3 4: T2 5: T1
Variable splits:
(1) split_variable: Longitude split_value: 84.953
(2) split_variable: Longitude split_value: 83.914
(4) split_variable: Latitude split_value: 25.426
(8) * action: 3
(9) * action: 5
(5) split_variable: nitrogen_0.5cm split_value: 2.1
(10) * action: 5
(11) * action: 4
(3) split_variable: wc2.1_30s_elev split_value: 51
(6) split_variable: Latitude split_value: 25.025
(12) * action: 3
(13) * action: 5
(7) split_variable: Longitude split_value: 84.964
(14) * action: 2
(15) * action: 5 plot(tr_sowing3)tr_sowing4 <- hybrid_policy_tree(X_cf_sowing, DR.scores_sowing, depth = 4)
tr_sowing4policy_tree object
Tree depth: 4
Actions: 1: T5 2: T4 3: T3 4: T2 5: T1
Variable splits:
(1) split_variable: Longitude split_value: 84.953
(2) split_variable: Longitude split_value: 83.914
(4) split_variable: Latitude split_value: 25.426
(8) split_variable: Longitude split_value: 83.913
(16) * action: 3
(17) * action: 1
(9) split_variable: Latitude split_value: 26.417
(18) * action: 5
(19) * action: 3
(5) split_variable: Longitude split_value: 83.929
(10) split_variable: Longitude split_value: 83.928
(20) * action: 4
(21) * action: 1
(11) split_variable: nitrogen_0.5cm split_value: 2.1
(22) * action: 5
(23) * action: 4
(3) split_variable: wc2.1_30s_elev split_value: 51
(6) split_variable: wc2.1_30s_elev split_value: 43
(12) split_variable: soc_5.15cm split_value: 14.5
(24) * action: 4
(25) * action: 5
(13) split_variable: Latitude split_value: 25.025
(26) * action: 3
(27) * action: 5
(7) split_variable: wc2.1_30s_elev split_value: 53
(14) split_variable: Longitude split_value: 85.415
(28) * action: 2
(29) * action: 5
(15) split_variable: Longitude split_value: 84.964
(30) * action: 2
(31) * action: 5 plot(tr_sowing4)tr_assignment_sowing=CSISA_KVKestim_sow
tr_assignment_sowing$depth2 <- predict(tr_sowing, X_cf_sowing)
table(tr_assignment_sowing$depth2)
2 4 5
38 200 3390 tr_assignment_sowing$depth3 <- predict(tr_sowing3, X_cf_sowing)
table(tr_assignment_sowing$depth3)
2 3 4 5
38 188 200 3202 tr_assignment_sowing$depth4 <- predict(tr_sowing4, X_cf_sowing)
table(tr_assignment_sowing$depth4)
1 2 3 4 5
40 47 168 304 3069 library(rgdal)
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2==1]="T5_16Dec"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2==2]="T4_15Dec"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2==3]="T3_30Nov"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2==4]="T2_20Nov"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2==5]="T1_10Nov"
tr_assignment_sowingsp= SpatialPointsDataFrame(cbind(tr_assignment_sowing$Longitude,tr_assignment_sowing$Latitude),data=tr_assignment_sowing,proj4string=CRS("+proj=longlat +datum=WGS84"))
library(mapview)
mapviewOptions(fgb = FALSE)
tr_assignment_sowingspmapview=mapview(tr_assignment_sowingsp,zcol="depth2_cat",layer.name="Recommended sowing dates")
tr_assignment_sowingspmapviewlibrary(ggridges)
library(dplyr)
tau.multi_sowing.forest=predict(multi_sowing.forest, target.sample = "all",estimate.variance=TRUE)
tau.multi_sowing.forest=as.data.frame(tau.multi_sowing.forest)
tau.multi_sowing.forest_X=data.frame(CSISA_KVKestim_sow,tau.multi_sowing.forest)
# Ridges -------------------
tau.multi_sowing.forest_pred=tau.multi_sowing.forest[,1:4]
library(dplyr)
library(reshape2)
tau.multi_sowing.forest_pred=rename(tau.multi_sowing.forest_pred,"T4_15Dec - T5_16Dec"="predictions.T4...T5.Y.1")
tau.multi_sowing.forest_pred=rename(tau.multi_sowing.forest_pred,"T3_30Nov-T5_16Dec"="predictions.T3...T5.Y.1")
tau.multi_sowing.forest_pred=rename(tau.multi_sowing.forest_pred,"T2_20Nov-T5_16Dec"="predictions.T2...T5.Y.1")
tau.multi_sowing.forest_pred=rename(tau.multi_sowing.forest_pred,"T1_10Nov-T5_16Dec"="predictions.T1...T5.Y.1")
tau.multi_sowing.forest_pred_long=reshape2::melt(tau.multi_sowing.forest_pred[,1:4])
ggplot(tau.multi_sowing.forest_pred_long, aes(x=value, y=variable, fill = factor(stat(quantile)))) +
stat_density_ridges(
geom = "density_ridges_gradient", calc_ecdf = TRUE,
quantiles = 4, quantile_lines = TRUE
) +
scale_fill_viridis_d(name = "Quartiles")+
theme_bw(base_size = 16)+labs(x="Wheat yield gain(t/ha)",y="Sowing date options")Warning: `stat(quantile)` was deprecated in ggplot2 3.4.0.
i Please use `after_stat(quantile)` instead.Warning: Using the `size` aesthetic with geom_segment was deprecated in ggplot2 3.4.0.
i Please use the `linewidth` aesthetic instead.
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2 == 1] <- "T5_16Dec"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2 == 2] <- "T4_15Dec"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2 == 3] <- "T3_30Nov"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2 == 4] <- "T2_20Nov"
tr_assignment_sowing$depth2_cat[tr_assignment_sowing$depth2 == 5] <- "T1_10Nov"
library(ggalluvial)
library(data.table)
tr_assignment_sowingDT = data.table(tr_assignment_sowing)
TransitionMatrix_sowing <- tr_assignment_sowingDT[, (sum <- .N), by = c("SowingSchedule", "depth2_cat")]
library(dplyr)
TransitionMatrix_sowing <- rename(TransitionMatrix_sowing, Freq = V1)
library(scales)
transitionmatrixplot_sowing <- ggplot(
data = TransitionMatrix_sowing,
aes(axis1 = SowingSchedule, axis2 = depth2_cat, y = Freq)
) +
geom_alluvium(aes(fill = depth2_cat)) +
geom_stratum() +
# geom_text(stat="stratum", aes(label=after_stat(stratum),nudge_y =5))+
geom_text(stat = "stratum", aes(label = paste(after_stat(stratum), percent(after_stat(prop))))) +
scale_x_discrete(
limits = c("SowingSchedule", "depth2_cat"),
expand = c(0.15, 0.05)
) +
scale_fill_viridis_d() +
theme_void(base_size = 20) +
theme(legend.position = "none")
transitionmatrixplot_sowingWarning: Using the `size` aesthetic in this geom was deprecated in ggplot2 3.4.0.
i Please use `linewidth` in the `default_aes` field and elsewhere instead.